Thanks a lot Emil! The patch has been commited to master (cd6efde).
-Raphael.
On 16.06.11 12:53, Emil Kroymann wrote:
OK,
apparently there is a second mapping from remote payload type to codec
already previewed in current SEMS master. However, because of two minor
bugs, this mapping wasn't used. Attached is a patch that fixes these
bugs. With this patch applied SEMS sends RTP with payload types
according to the mapping given in the remote SDP.
Regarding the "SHOULD" in the RFC reported by Matthew: This can be fixed
by introducing a second "local" mapping from payload type to codec per
RTP stream. However, I currently have no time to implement such a
thing, since the behavior without the should is acceptable for us at
the moment.
Regards,
Emil
Am Mon, 13 Jun 2011 14:44:54 +0200
schrieb Emil Kroymann<[email protected]>:
Hi Raphael,
IMHO we need two mappings from payload type two codec per RTP session
anyway, since when SEMS is generating the offer, the remote end can
answer with any mapping it likes and SEMS must use this mapping for
sending. If the payload type mapping in SEMS is not global anymore, we
can adjust the mapping in the SDP answer generated by SEMS to whatever
is given in the offer.
Regards,
Emil
Am Sun, 12 Jun 2011 08:39:31 +0200
schrieb Raphael Coeffic<[email protected]>:
Hi Emil, Matthew,
I'm quite sure how to realize such a "remote bugfix", as it could
overlap with locally define codecs...
By the way, I've seen this behavior with Jitsi (aka. SIP
communicator). However, my snom 360 behaves correctly.
Neither of the other UAs I have ever used do expose such a
missbehavior.
Matthew: were you using Jitsi (possibly the osx versiob ;-))? If
yes, we should definitely file a bug report.
Cheers
Raphael.
On 11.06.11 17:53, Emil Kroymann wrote:
Hi Matthew,
I think in this case it is the peer behaving incorrectly. The peer
should send under the payload types offered by SEMS and SEMS
vice-versa under the payload types given in the answer.
Still it seems to be desirable for interoperability, if SEMS would
match the payload types given by the peer. In the case where SEMS
is generating the SDP offer, this is not possible, however.
Regards,
Emil
Am Sat, 11 Jun 2011 11:02:10 -0400
schrieb Matthew Williams<[email protected]>:
Hi,
I think I'm finding a very similar issue with payload types for
telephone-events. If this is unrelated, please accept my
apologies.
A peer is sending an INVITE with a payload type of '101' for
telephone-events. In SEMS' reply (200 OK), payload type '96' is
indicated for telephone-events. When the peer sends an event with
type 101, SEMS rejects it as unknown type. I was actually
expecting SEMS to reply with the same payload type (101), but
from what you're saying, it seems the sdp itself is acceptable,
but the fact that SEMS is rejecting payload type 101 is not? Or
is it the peer that should be sending as type 96, since this is
what was in the SDP reply?
I am working with latest master
(856a7428eaea24a426688b277b2d504b4cb1f798).
Regards,
Matthew Williams
On Sat, Jun 11, 2011 at 7:05 AM, Emil Kroymann
<[email protected]>wrote:
Hi,
I think I found a problem with the SDP offer-answer
implementation recently merged into master while implementing
wideband functionality. The problem is with the assignement of
dynamic payload types. SEMS seems to keep only one mapping of
dynamic payload types to codec implementation, which is global
to all of SEMS. However, it should keep two separate mappings -
one for sending and one for receiving - for each RTP stream.
This problem shows in the following setup:
I use twinkle to setup a call to SEMS, which plays an
announcement using the speex wideband codec. Twinkle sends the
following SDP in the INVITE:
v=0.
o=twinkle 1229564934 1632772146 IN IP4 192.168.1.121.
s=-.
c=IN IP4 XX.XX.XX.XX.
t=0 0.
m=audio 13006 RTP/AVP 98 97 8 0 102 3 101.
a=rtpmap:98 speex/16000.
a=rtpmap:97 speex/8000.
a=rtpmap:8 PCMA/8000.
a=rtpmap:0 PCMU/8000.
a=rtpmap:102 G726-16/8000.
a=rtpmap:3 GSM/8000.
a=rtpmap:101 telephone-event/8000.
a=fmtp:101 0-15.
a=ptime:20.
a=nortpproxy:yes.
This means, that twinkle expects the speex wideband codec under
payload type 98 and the speex narrowband codec under payload
type 97.
SEMS answers with the following SDP in the 200 OK:
v=0. o=sems 1 1 IN IP4 XX.XX.XX.XX.
s=sems.
c=IN IP4 89.246.236.49.
t=0 0.
m=audio 10000 RTP/AVP 97 96 8 0 101.
a=rtpmap:97 speex/16000.
a=rtpmap:96 speex/8000.
a=rtpmap:8 PCMA/8000.
a=rtpmap:0 PCMU/8000.
a=rtpmap:101 telephone-event/8000.
This means, that SEMS expects the speex wideband codec under
payload type 97 and the speex narrowband codec under payload
type 96.
Unfortunately, SEMS sends speex wideband encoded RTP packets
with payload type 97. This leads to twinkle decoding the audio
with the narrowband decoder and to poor audio quality.
As mentioned above, to fix the problem, SEMS has to keep a
seperate payload type mapping for sending and it has to set the
payload type on outgoing packets according to this mapping.
Since we need this functionality at ISACO quite urgently, I
would be willing to implement a fix for this.
What do you think?
Regards,
Emil
--
Emil Kroymann
VoIP Services Engineer
Email: [email protected]
Tel: +49-30-203899885
Mobile: +49-176-38389303
ISACO GmbH
Kurfürstenstraße 79
10787 Berlin
Germany
Amtsgericht Charlottenburg, HRB 112464B
Geschäftsführer: Daniel Frommherz
_______________________________________________
Sems mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/sems
_______________________________________________
Sems mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/sems
_______________________________________________
Sems mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/sems
_______________________________________________
Sems mailing list
[email protected]
http://lists.iptel.org/mailman/listinfo/sems
